Two researchers at USC’s Information Sciences Institute used a large language model pipeline to classify all 58,746 NIH-funded projects listed for fiscal year 2025. Their portfolio analysis identified 9,363 projects with a substantive AI or machine learning component, then mapped their funding, disease areas, development stages, methods, and institutional collaborations. The results show where NIH-backed AI work is concentrated and how little of it has reached clinical implementation.

One in six projects uses AI

AI-related work accounted for 15.9% of the NIH projects studied. Those projects received an average award of $675,129, which was 13.4% higher than the $595,135 average for projects outside the AI classification.

Cancer and oncology represented the largest AI research area, followed by aging and neurodegenerative disease, mental health and substance use disorders, neuroscience, and infectious disease.

Health disparities and minority health research received 5.7% of AI funding, despite its status as an NIH priority.

Charts showing AI adoption, funding, research stages, and disease areas across the NIH portfolio
AI adoption and funding across the fiscal year 2025 NIH portfolio.

Clinical deployment gets a narrow slice

Most AI projects remained in research or development. The authors classified 7,393 projects, or 79.0%, in that stage. Clinical deployment or implementation accounted for 1,372 projects, or 14.7%.